Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval
Convolutional neural networks (CNNs) are frequently utilized in content-based remote sensing image retrieval (CBRSIR). However, the features extracted by CNNs are not rotationally invariant, which is problematic for remote sensing (RS) images where objects appear at variable rotation angles. In addi...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2024-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10483252/ |
_version_ | 1797201331136495616 |
---|---|
author | Zhoutao Cai Yukai Pan Wei Jin |
author_facet | Zhoutao Cai Yukai Pan Wei Jin |
author_sort | Zhoutao Cai |
collection | DOAJ |
description | Convolutional neural networks (CNNs) are frequently utilized in content-based remote sensing image retrieval (CBRSIR). However, the features extracted by CNNs are not rotationally invariant, which is problematic for remote sensing (RS) images where objects appear at variable rotation angles. In addition, because RS images contain a wealth of content and detailed information, CNNs may lead to information loss by superimposing multiple convolutional and pooling layers, affecting the ability of the model to extract features. To address these problems, this article proposes a proxy-based feature fusion network. By designing a proxy-based Euclidean distance contrast loss that combines contrast learning within the framework of metric learning, such that the distance between the source image and its rotated image embedding vector in the metric space is closer than any other image, thus endowing the model with a certain degree of rotation invariant. Meanwhile, the global correlation map is generated by multilayer fusion, under whose guidance the features of each layer are fused to improve the feature extraction capability of the model and to reduce the loss in the image flow process. Extensive experiments based on two public RS datasets show that the method achieves better performance compared to other methods. |
first_indexed | 2024-04-24T07:45:50Z |
format | Article |
id | doaj.art-83caadf994014ad691424e9d8cc91d14 |
institution | Directory Open Access Journal |
issn | 1939-1404 2151-1535 |
language | English |
last_indexed | 2024-04-24T07:45:50Z |
publishDate | 2024-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-83caadf994014ad691424e9d8cc91d142024-04-18T23:00:18ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352024-01-01177759777210.1109/JSTARS.2024.338284510483252Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image RetrievalZhoutao Cai0https://orcid.org/0009-0001-8764-0232Yukai Pan1https://orcid.org/0009-0005-2694-2244Wei Jin2https://orcid.org/0000-0002-6844-4324Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaConvolutional neural networks (CNNs) are frequently utilized in content-based remote sensing image retrieval (CBRSIR). However, the features extracted by CNNs are not rotationally invariant, which is problematic for remote sensing (RS) images where objects appear at variable rotation angles. In addition, because RS images contain a wealth of content and detailed information, CNNs may lead to information loss by superimposing multiple convolutional and pooling layers, affecting the ability of the model to extract features. To address these problems, this article proposes a proxy-based feature fusion network. By designing a proxy-based Euclidean distance contrast loss that combines contrast learning within the framework of metric learning, such that the distance between the source image and its rotated image embedding vector in the metric space is closer than any other image, thus endowing the model with a certain degree of rotation invariant. Meanwhile, the global correlation map is generated by multilayer fusion, under whose guidance the features of each layer are fused to improve the feature extraction capability of the model and to reduce the loss in the image flow process. Extensive experiments based on two public RS datasets show that the method achieves better performance compared to other methods.https://ieeexplore.ieee.org/document/10483252/Deep neural networkfeature fusionmetrics learningremote sensing (RS) image retrievalrotation invariant |
spellingShingle | Zhoutao Cai Yukai Pan Wei Jin Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Deep neural network feature fusion metrics learning remote sensing (RS) image retrieval rotation invariant |
title | Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval |
title_full | Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval |
title_fullStr | Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval |
title_full_unstemmed | Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval |
title_short | Proxy-Based Rotation Invariant Deep Metric Learning for Remote Sensing Image Retrieval |
title_sort | proxy based rotation invariant deep metric learning for remote sensing image retrieval |
topic | Deep neural network feature fusion metrics learning remote sensing (RS) image retrieval rotation invariant |
url | https://ieeexplore.ieee.org/document/10483252/ |
work_keys_str_mv | AT zhoutaocai proxybasedrotationinvariantdeepmetriclearningforremotesensingimageretrieval AT yukaipan proxybasedrotationinvariantdeepmetriclearningforremotesensingimageretrieval AT weijin proxybasedrotationinvariantdeepmetriclearningforremotesensingimageretrieval |